Liat Ayalon, Department of Psychiatry 151B, University of California, San Diego, 3350 La Jolla Village Dr., San Diego, CA 92161, USA. Tel.: 858-552-8585 ext. 2643; fax: 858-552-7536; e-mail: email@example.com
This study examined response inhibition during a Go–NoGo task in individuals with obstructive sleep apnea (OSA). Fourteen OSA patients and 14 controls were studied with functional magnetic resonance imaging. Compared to controls, the OSA group showed more false positives (error of commission) during the NoGo trials with decreased brain activation in the left postcentral gyrus, cingulate gyrus and inferior parietal lobe, as well as right insula and putamen. This is consistent with previous findings of impaired performance and decreased brain activation in OSA patients during a working memory task, suggesting that compromised brain function in response to cognitive challenges may underlie some of the cognitive deficits seen in patients with OSA.
Obstructive sleep apnea (OSA) syndrome is characterized by repeated episodes of upper airway obstruction during sleep that result in intermittent hypoxemia and arousals (Malhotra and White, 2002) and occurs in 4–9% of middle-aged men and women (30–60 years), and 45–62% of older adults (60+ years) (Ancoli-Israel et al., 1991; Young et al., 2002). Individuals with OSA commonly present with impairments in cognition including various aspects of executive function, psychomotor speed and attention (Beebe et al., 2003; El Ad and Lavie, 2005).
Despite the considerable data about the behavioral correlates of OSA, relatively little is known about the neural substrates underlying the behavioral effects. Thomas et al. (2005) reported impaired performance on a working memory task in individuals with OSA compared to controls, along with reduced activation within the anterior cingulate, dorsolateral prefrontal cortex and posterior parietal cortex in the OSA group. Our group (Ayalon et al., 2006) examined performance and cerebral responses during a verbal learning task in OSA and controls and found the opposite findings: increased brain activation and intact immediate free recall performance in the OSA group compared to controls.
These two studies suggest that alterations in cerebral responses to cognitive challenges in OSA may depend on the cognitive domain assessed. This generalization is limited, however, by the fact that only two cognitive domains have been reported in the literature. Here we compared cerebral activation during a response inhibition task (the Go–NoGo task) in OSA and controls using functional magnetic resonance imaging (FMRI). The Go–NoGo task involves frequent automatic responding to stimuli interspersed with the need to withhold (i.e. inhibit) a response to a specific, less frequently occurring, stimulus. Our focus here was on the response inhibition component: the need to stop oneself from engaging in a prepotent response when that reaction is not appropriate.
Given that OSA patients show behavioral impairment in two cognitive components involved in response inhibition (attention to incoming stimuli and prevention of an automatic response; Lezak et al., 2004) (Beebe et al., 2003; Quan et al., 2006), we hypothesized that individuals with OSA, relative to good sleepers, would show impaired performance on the Go–NoGo task and decreased cerebral activation in task-related brain regions.
Thirty-eight participants signed informed consent and underwent a screening overnight polysomnography (PSG). For inclusion, OSA patients needed an apnea–hypopnea index (AHI) ≥10 and controls needed an AHI <5. We excluded two prospective patients who had sleep disorders other than OSA, five controls who had an AHI >5 or other sleep disorders, and three OSA patients who could not tolerate the MRI due to their girth. In the end, 28 subjects (14 treatment-naïve OSA patients and 14 matched healthy controls; 13 male, one female in each group) participated in the FMRI, all of whom provided usable imaging data. There were no significant differences between the patients and controls on any demographic variable (Table 1). All participants were right-handed, healthy and free of current and past psychiatric and medical disorders, as determined by history and physical exam, the Composite International Diagnostic Interview for mental disorders, routine lab work and urine toxicology screens. We excluded individuals with hypertension (>180/110), diabetes and body weight over 300 pounds. The study was approved by the local IRB, and all participants provided written informed consent.
Table 1. Sample characteristics
OSA (n = 14)
Control (n = 14)
Values are given as mean (SD).
P-value is for independent samples t-test.
AHI, apnea–hypopnea index; BMI, body-mass index; OSA, obstructive sleep apnea.
Blood pressure (systolic)
Blood pressure (diastolic)
The protocol consisted of a screening appointment and overnight PSG 5–14 days later followed by an FMRI scan the next morning. PSG was used to confirm OSA/control diagnosis and rule out sleep disorders other than OSA. Standard electroencephalograms, electrooculogram, chin electromyogram, electrocardiogram, airflow, thoracic and abdominal excursions, oximetry and tibialis electromyogram were recorded. Apnea was defined as any ≥10 s of ≥80% drops of respiratory amplitude. Hypopnea was defined as any ≥10 s of ≥30% drops of respiratory amplitude, plus ≥3% desaturation or arousal. AHI was calculated as the number of apnea and hypopnea events per hour of sleep.
Two to three hours after waking on the morning after the PSG, participants underwent an FMRI. During the FMRI session, participants performed the Go–NoGo response inhibition task, as well as a short-term memory and a sensorimotor task (results from these tasks are reported elsewhere; Ayalon et al., 2006).
The Go–NoGo task alternated between active blocks and resting blocks. During resting blocks, participants viewed a fixation cross in the center of the screen. During active blocks, a large square, small square, large pentagon and small pentagon were presented one at a time in the center of the screen. Stimuli appeared for 200 ms every 1500 ms. Participants were instructed to press a button as fast as possible every time they saw a shape (Go stimuli) but to withhold a response when they saw the small pentagon (the NoGo stimulus). The task lasted 6:24 min. Functional images consisted of 126 gradient echo-planar images (time repetition: 3 s; echo time: 30 ms; field of view: 256 mm; resolution = 4 × 4 × 4 mm) of 32 axial slices covering the whole brain and measuring the blood oxygenation level-dependent (BOLD) signal. Functional data were aligned with high-resolution anatomical images (fast spoiled gradient recalled echo: 1 mm3 resolution).
At the conclusion of the task, while still in the scanner, participants were administered the Stanford Sleepiness Scale (Hoddes et al., 1973) and Karolinska Sleepiness Scale (Akerstedt and Gillberg, 1990) and 10-point Likert scales assessing the following subjective factors: task difficulty, ability to concentrate, effort put into the task and motivation to perform the task well.
Performance and postscan questionnaire data were analyzed with Student's t-tests (equal variances not assumed) comparing the OSA and the control groups. Response inhibition was measured with false positive rate (i.e. errors of commission for the NoGo shape).
Functional magnetic resonance imaging data were processed and analyzed with afni (Cox, 1996). Individual time-course BOLD data were analyzed with a general linear model (Ward, 2002) with the following parameters: constant, linear drift, six motion correction parameters and two reference functions: one comparing Go stimuli to rest blocks and one comparing NoGo stimuli to rest blocks, each convolved with an idealized hemodynamic response function (Cohen, 1997). The parameter used in group analyses was the regression coefficient associated with the NoGo stimuli. Prior to group analyses, data sets were smoothed with a Gaussian filter of 4.0 mm full-width half-maximum and transformed to standard atlas coordinates (Talairach and Tournoux, 1988).
We determined group differences in the cerebral response to inhibition using independent samples t-tests. To assess the correlation between brain activation and performance in the OSA group, BOLD response data were regressed onto performance data. For the performance analysis, a search region approach (Eyler Zorrilla et al., 2003) was taken to confine the analysis to areas showing significant group differences. A regression was conducted examining the relationship between performance and activation in brain regions, showing significant group differences in the initial analysis above.
For all FMRI analyses, we utilized a whole brain voxel-wise analysis. To protect against Type I error, we used a cluster threshold method (Forman et al., 1995). This required any given voxel to be statistically significant at the P < 0.05 level and part of a cluster of at least 12 contiguous voxels (768 mm3), each individually significantly activated. Hence, the clusters we report are equal to or larger than the single largest cluster of activation expected by chance at α = 0.05.
Behavioral data for one OSA patient were missing due to equipment failure. Although not statistically significant, the OSA group had almost double the false positive rate (error of commission) compared to the control group during the NoGo trials (OSA: mean = 11%, SD = 12%; control: mean = 6%, SD = 3%; t(25) = 1.32, P = 0.20) (Glass’s effect size = 1.29). Both groups reported similar ability to concentrate, perceived task difficulty, amount of effort put into the task and motivation to perform well. Differences between the groups in subjective sleepiness were also not significant (Table 2).
Table 2. Subjective sleepiness and post-scan data
OSA (n = 14)
Control (n = 14)
Values are given as mean (SD).
OSA, obstructive sleep apnea.
Stanford Sleepiness Scale
Karolinska Sleepiness Scale
Ability to concentrate
Perceived task difficulty
Amount of effort put into the task
Motivation to perform well
In the control group, the NoGo trials activated brain regions typically activated during performance of response inhibition tasks, including bilateral cingulate [Brodmann’s areas (BA) 24 and 32] and medial frontal gyrus (BA6), left precentral gyrus (BA4), inferior occipital gyrus (BA19), putamen and right insula.
The OSA group showed decreased brain activation in several brain regions compared to controls during the NoGo trials. These regions included left postcentral gyrus, cingulate gyrus and inferior parietal lobe, as well as right inferior frontal gyrus (BA47) and insula, and right putamen (Table 3, Fig. 1).
Table 3. Brain regions showing group differences (obstructive sleep apnea versus control) during the NoGo trials
Center (X, Y, Z)
Clusters shown survived our cluster threshold alpha protection procedure (whole brain P < 0.05; see Methods for details).
Anatomical locations and Brodmann’s areas (BA) list every region covered by a given cluster. This does not imply that all portions of a listed region were covered.
L, left; R, right; eta2, effect size measuring variance accounted for by group.
Inferior frontal and insula
34, 18, 3
L32/24 and R24
−1, 7, 36
−52, −31, 36
−19, −47, 64
−55, −19, 17
24, 7, −3
In terms of the relationship with performance in the OSA group, a higher false positive rate (i.e. greater disinhibition) was associated with increased activation in the left caudal zone of the anterior cingulate (BA24) (volume = 256; center coordinates: X = −6, Y = 4, Z = 40; effect size: R2 = 0.55).
The findings suggest that individuals with OSA showed impaired cerebral activation during response inhibition relative to controls. In particular, the OSA group showed diminished response in regions typically involved with response inhibition (Simmonds et al., 2008), conflict monitoring (cingulate), attention inferior parietal lobule, motor function (postcentral gyrus, putamen) and decision making (insula) (Drummond et al., 2005; Garavan et al., 2006; Hester et al., 2004; Wager et al., 2005). We also found higher false positive rates in OSA compared with controls; however, these differences were not statistically significant. This may be due to our relatively small sample size and the large variability in performance in the OSA group. In addition, FMRI may be more sensitive in detecting differences between the groups than performance measure, and changes in brain activation may predict a subsequent decline in cognition (Bookheimer et al., 2000).
The observed positive relationship between errors of commission and activation in the caudal zone of the anterior cingulate is consistent with reports of involvement of this area in error processing (Stevens et al., 2007). Thus, the increased activation here may reflect subjects’ awareness of having made an error.
The data reported here are consistent with the finding of decreased cerebral activation during a working memory task in individuals with OSA (Thomas et al., 2005), suggesting that individuals with OSA show impaired cerebral responses during tasks traditionally thought to rely on some component of executive function. The findings are also consistent with findings of impaired response inhibition and reduced task-related activation, including lowered Go–NoGo-related activation, in healthy individuals after sleep deprivation (relative to normal sleep) (Chee et al., 2006; Chuah et al., 2006). One notable consequence of OSA is sleep fragmentation, thus sleep fragmentation may be contributing to some of the cognitive deficits in OSA patients. This interpretation would be consistent with experimental work in sleep fragmentation showing deficits similar to sleep deprivation (Bonnet and Arand, 2003).
In contrast to the findings here, we previously reported (using a subset of n = 12/group from this sample) intact verbal short-term memory performance associated with increased cerebral activation in OSA compared to controls (Ayalon et al., 2006). As greater cerebral response within OSA patients was associated with better memory performance, we suggested that recruitment of additional brain regions to participate in verbal encoding in individuals with OSA likely represents an adaptive compensatory recruitment response. Given the results from these two studies and the third one in the literature (Thomas et al., 2005), such compensatory recruitment would appear related to the specific cognitive demands of a given task, suggesting that OSA may differentially impact various brain regions and/or cognitive processes.
Although our data cannot conclusively demonstrate that differences in baseline physiology [e.g. cerebral blood flow (CBF)] are not partially responsible for altered activation in the OSA group, we minimized these possible confounds by excluding participants with conditions that perturb basal CBF and/or cerebral neurovascular coupling and by matching the groups on age, BMI and blood pressure. Furthermore, the fact that these same OSA patients showed no differences from controls on a sensorimotor task and increased activation on a verbal encoding task (previously reported in Ayalon et al., 2006) suggests that the observed changes in brain activation in response to cognitive challenges do not reflect a global change in the baseline BOLD signal.
Clearly, the exclusion of individuals weighing >300 pounds (due to MRI weight limit), and individuals with hypertension or diabetes, necessarily limits the generalizibility of the findings to the overall OSA population.
Taken together, studies using FMRI suggest that individuals with OSA differ from those without OSA in their cerebral response to cognitive challenges. Importantly, cognitive task-related factors may have an effect on whether an increased or decreased response to cognitive challenges is observed. The decrease in brain activation reported here using a response inhibition task and by Thomas et al. (2005) using a working memory task may be related to executive functioning deficits often associated with OSA (Beebe et al., 2003). On the other hand, increase in brain activation during a verbal encoding task in individuals with OSA (Ayalon et al., 2006) may be related to the relatively preserved verbal and short-term memory functions in this population (Beebe et al., 2003). Further studies with larger sample size are needed to confirm the results reported here and elsewhere. Additionally, future work should examine a wider range of cognitive tasks and measures of brain function (including CBF), as well as the influence of OSA clinical measures, to fully characterize brain dysfunction in OSA.
Support: NIH M01 RR00827, NIMH 5 T32 MH18399, NIA AG08415, National Sleep Foundation Pickwick Fellowship (LA).